| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Volen Center, Brandeis University, Waltham, Massachusetts
Submitted 11 January 2008; accepted in final form 25 March 2008
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
). A large number of studies have focused on when and how this spontaneous network activity is generated and possible roles this activity plays in the proper development of the network (Ben-Ari 2001; Cang et al. 2005; Demas et al. 2006; Fischer et al. 1998; Garaschuk et al. 1998; Greer et al. 2006; Gu and Spitzer 1995; Hanson and Landmesser 2003, 2004, 2006; Hooks and Chen 2006; Huberman et al. 2006; Leinekugel et al. 2002; Mehta and Sernagor 2006; Milner and Landmesser 1999; Moody and Bosma 2005; O'Donovan 1999; Penn et al. 1998; Ren and Greer 2003; Sernagor et al. 2001; Stellwagen and Shatz 2002; Torborg and Feller 2005; Warland et al. 2006).
Studies in the mammalian retina (Demas et al. 2003; Warland et al. 2006; Wong and Oakley 1996; Wong et al. 1993), spinal cord (Bekoff 1976; Nishimaru and Kudo 2000; Ren and Greer 2003; Sharp et al. 1999), and brain stem (Greer et al. 2006; Pagliardini et al. 2003) have documented the age-related progression of changes from early spontaneous activity to adult-like activity. There are common trends in the maturation of activity patterns: embryonic spontaneous network activity is often episodic and characterized by synchronous firing of many neurons, and this synchronous activity can begin to disappear before eye opening, walking, or breathing. There does not seem to be a universal program, however, for the transition from early spontaneous network activity to mature activity. Spontaneous activity diminishes at different rates along the rat spinal cord (Ren and Greer 2003), whereas in the mouse spinal cord, spontaneous activity disappears at similar times along the spinal cord, but more mature motor patterns appear at different rates along the spinal cord (Yvert et al. 2004). There are also many kinds of transitional activity in different networks. In the mouse retina (Demas et al. 2003), transitional activity is distinct from both the immature and mature activity patterns, but in the spinal cord of the metamorphosing frog, the transitional state includes both immature and mature forms of activity (Combes et al. 2004).
In developing motor systems, one can ask a number of questions. Do mature rhythms appear suddenly or gradually? Do mature and immature motor patterns occur simultaneously in the same neuron? Is there a reliable temporal structure to this transition? Do mature patterns appear in an organized pattern over time, or do they emerge in a more stochastic manner? What is changing in the network and/or inputs to the network that could explain the appearance of mature activity patterns?
The decapod crustacean stomatogastric nervous system (STNS) is a useful system for the study of network maturation because the rhythmic movements of the adult stomach are stereotyped and the neuronal network that controls these motor patterns is relatively small—
30 neurons are contained within the stomatogastric ganglion (STG)—and has been well characterized (Harris-Warrick et al. 1992). The STNS controls a number of rhythmic movements of the stomach. One of these rhythms, the gastric mill rhythm, controls the movements of a set of teeth within the foregut that chews food in the stomach. The adult gastric mill rhythm typically has a frequency of 0.1 Hz. Another rhythm, the pyloric rhythm, controls the movements of the pylorus, which filters food particles of different sizes. The frequency of the adult pyloric rhythm is typically 1 Hz. Despite the difference in frequency, these two rhythms are not completely independent. For example, adult motor neurons produce activity reflecting both rhythms at the same time (Bucher et al. 2006; Heinzel 1988; Heinzel and Selverston 1988; Mulloney 1977; Thuma and Hooper 2002; Weimann et al. 1991).
Before hatching, and therefore before feeding, the embryonic STNS generates a rhythm in which the gastric mill and pyloric neurons are all active at about the same frequency (Casasnovas and Meyrand 1995). This has been interpreted as implying that embryonic gastric mill and pyloric neurons are part of a unified rhythmic pattern generator. Fully differentiated and recognizable pyloric and gastric mill rhythms are not clearly evident until metamorphosis from the third larval stage to the first stage of postlarval growth (Casasnovas and Meyrand 1995).
However, does this imply that no features of the gastric mill network are present early in development? Intriguingly, a previous report briefly mentioned that gastric-timed activity was occasionally observed in the output from embryonic gastric mill neurons (Casasnovas and Meyrand 1995). However, to date no attempt has been made to quantify the extent to which gastric mill and pyloric neurons show activity at different frequencies in the embryonic network. To this end, we used spectral methods related to those used previously by Bucher et al. (2006) to analyze long recordings of the output from multiple gastric mill and pyloric neurons. Our results confirmed the earlier finding that embryonic neurons share a common unified frequency, but we also found that gastric mill neurons reliably produced intermittent slow activity reminiscent of the adult gastric mill rhythm. This suggests that an embryonic precursor of the gastric mill rhythm is present even at this early stage of development.
Further support for this idea came from the results of applying a modulator, Cancer borealis tachykinin-related peptide 1a (CabTRP), to the embryonic STNS. CabTRP is a neuropeptide found in the neuromodulatory inputs to the STG in adult crabs and lobsters (Blitz et al. 1995; Christie et al. 1997; Goldberg et al. 1988; Thirumalai and Marder 2002). In crabs, CabTRP is found in one of the neurons that activates gastric mill activity (Nusbaum et al. 2001; Wood et al. 2000). Previous work in the isolated STG (with modulatory inputs removed) showed that CabTRP strongly activated some, but not all, neurons in the pyloric network of the lobster, Homarus americanus (Thirumalai and Marder 2002). However, the effects of CabTRP on the motor patterns generated by the intact lobster STG (modulatory inputs present) have not been reported. Furthermore, CabTRP appears in the neuromodulatory inputs to the lobster STG late in larval development (Cape et al. 2008; Fénelon et al. 1999). Therefore we compared the effects of CabTRP on the adult and embryonic STNS to ask whether CabTRP would activate gastric mill rhythms similarly at both developmental stages. We found that CabTRP increased the amount of slow activity in gastric mill neurons at both stages. Thus the immaturity of the embryonic neuromodulatory inputs is likely to contribute to the immaturity of the output from the embryonic STNS. Taken together, these results suggest that the STG network can express adult-like rhythmic behavior before fully differentiated adult motor patterns are observed, and that the maturation of neuromodulatory inputs is likely to play a role in the eventual establishment of the adult motor patterns.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
550–700 g) were purchased from Yankee Lobster (Boston, MA). Embryonic lobsters were harvested from three different egg-bearing lobsters obtained from Dr. M. Tlusty at the New England Aquarium (Boston, MA) and M. Syslo at the Massachusetts State Lobster Hatchery (Martha's Vineyard, MA). Embryos were removed from their egg cases and staged using an eye index according to Helluy and Beltz (1991). Embryos in control saline experiments were 70–99% of embryonic development (E70–E99) and embryos used for the CabTRP experiments were E78–E98. Animals were kept in recirculating artificial seawater tanks at 8–16°C. When indicated, 10–6 M CabTRP 1a (gift from Dr. Michael P. Nusbaum, synthesized at the Protein Chemistry Laboratory, University of Pennsylvania, Philadelphia, PA) was bath applied at 0.05 ml/s using a continuously flowing superfusion system. Electrophysiological recordings
Adult lobsters were anesthetized on ice for 10–20 min. The stomatogastric nervous system was removed from the stomach and pinned in a Sylgard-coated petri dish and superfused with chilled (10–14°C) saline. Extracellular recordings were obtained using stainless steel pin electrodes placed in petroleum jelly wells around the nerves that contain axons of particular gastric and pyloric neurons: dorsal gastric nerve (dgn) for dorsal gastric (DG); lateral gastric nerve (lgn) for lateral gastric (LG); pyloric dilator nerve (pdn) for pyloric dilator (PD); lateral pyloric nerve (lpn) for lateral pyloric (LP) and medial gastric (MG) (Bucher et al. 2006). Intracellular recordings from identified somata in the STG were made with glass microelectrodes (20–30 M
).
The whole embryo was removed from its egg case and pinned ventral side down in a plastic petri dish coated with a 0.5-cm-thick layer of transparent Sylgard (Dow Corning, Midland, MI; http://www.dowcorning.com) with 0.001-gauge tungsten wire (California Fine Wire Company, Grover Beach, CA; http://www.calfinewire.com). The carapace and yolk were removed, and a lateral incision was made up the side of the stomach, preserving the PD innervated muscles. The stomach was pinned flat, ventral side down, with 0.0005-gauge tungsten wire and superfused with chilled (10–16°C) saline. It is impossible to remove the STNS from the embryonic stomach; therefore output from embryonic neurons was obtained by intracellular recordings of excitatory junctional potentials (EJPs) from muscles innervated by specific pyloric and gastric mill neurons. Identifications of muscles were made according to Richards (2000). Glass microelectrodes (20–40 M) were used to obtain muscle recordings. LP neuron activity was recorded from either cardio-pyloric valve muscle 6 (cpv6) or pyloric muscle 1 (p1), both of which are innervated by the LP neuron. These two muscles are referred to collectively as lpm. PD neuron activity was recorded from cardio-pyloric valve muscle 2 (cpv2), referred to as pdm; DG neuron activity in gastric mill muscle 4 (gm4), referred to as dgm; LG neuron activity in gastric mill muscle 6 (gm6) or gastric mill muscle 8 (gm8), which we refer to collectively as lgm; and MG neuron activity in gastric mill muscle 9 (gm9), referred to as mgm. Simultaneous recordings from neighboring muscles were made to confirm muscle identification and boundaries between muscles. Recordings ranged from 10 to 110 min (mean = 38 min; median = 33.5 min). In 20% of the recordings, short periods of time (10 s to several minutes) were removed from the analysis because one of the electrode penetrations was lost and recovered. None of these recordings is shown in the figures.
The saline composition was (in mM) 479.12 NaCl, 12.74 KCl, 13.67 CaCl2, 20 Mg2SO4, 3.91 Na2SO4, and 5 HEPES; pH, 7.4–7.5. The intracellular microelectrode solution composition was 0.6 M K2SO4 and 20 mM KCl. Signals were amplified by Axoclamp 2B amplifiers (Axon Instruments/Molecular Devices, Sunnyvale, CA; http://www.moleculardevices.com) and digitized using a DigiData 1200 data acquisition board (Axon Instruments/Molecular Devices).
Data analysis
Recordings were analyzed with Spike2 (version 5, CED, Cambridge, UK; http://www.ced.co.uk) and MATLAB (version 7, The MathWorks, Natick, MA; http://www.mathworks.com). Statistical tests and plots were performed in MATLAB and Statview (version 5.0.1, SAS Institute, Cary, NC; http://www.sas.com). Figures were created in Canvas (version 10, ACD Systems, Miami, FL; http://www.acdsee.com).
Statistics
All error bars represent SE. One-way ANOVAs with Fisher's PLSD post hoc tests were used to compare the mean normed amplitudes in control saline. Unpaired t-test and one-way ANOVAs were performed to compare fundamental frequencies and the normed fast and slow amplitudes of the output from two or three neurons, respectively. Unpaired t-tests were used to compare the fundamental frequency and the amount of fast and slow activity in embryos and adults. The Watson-Williams test (Mardia and Jupp 1999) (Igor Pro 6.02A, WaveMetrics, Lake Oswego, OR; http://www.wavemetrics.com) was used to compare the coherence phase angles in adults and embryos for each cell type for the control saline experiments. Paired t-tests were used to compare the fundamental frequencies in control and CabTRP. To test whether changes in coherence phase angle caused by CabTRP application were significant, we used a circular analog of the paired t-test (Mardia and Jupp 1999; Upton 1973).
Spectral analysis
The adult STG generates two distinct rhythms, as seen in the extracellular nerve recordings in Fig. 1 A. The faster pyloric rhythm, which typically has a frequency of 1 Hz, is expressed by the PD and LP neurons, as seen in the nerve recordings pdn and lpn, respectively. The gastric mill rhythm, typically 0.1 Hz, is seen as activity in the DG (dgn), LG (lgn), and MG (lpn) neurons (Fig. 1A). A cell can participate in both rhythms simultaneously, resulting in an activity pattern that includes both fast (pyloric) and slow (gastric) bursting. This can be seen in the recordings of the adult DG and LG neurons, in which there are pyloric-timed events within each slow gastric burst (Fig. 1A).
|
40 ms. Spurious threshold crossings were eliminated by deleting any crossings that occurred <20 ms after the previous crossing. The spike trains were converted to a continuous rate-versus-time signal, r(t), by filtering the spike train (interpreted as a train of impulses) with a Gaussian filter (
= 50 ms). The continuous signal r(t) was "sampled" at a rate of 250 Hz. These signals appeared by eye to be a reasonable representation of the gastric-like and pyloric-like modulation of the spike rate.
To examine how the frequency content of the signals varied over time, we calculated spectrograms of the signals by dividing the signal into a number of temporal windows and estimating the power spectrum for each window (Thomson 2000). The power spectrum is a representation of how much of the variation in a signal can be accounted for by variation at different frequencies and is equal to the Fourier transform of the autocorrelation function of the signal. When a large amount of the variation in a signal can be accounted for by variation at a particular frequency, there is a lot of "power" at that frequency. The spectrogram therefore shows which frequencies contribute to a signal over time.
We estimated the power spectrum for each window using multitaper spectral estimation (Percival and Walden 1993), using windows of length T 
160 s (range, 138.9–169.7 s) for the control saline experiments and T 100 s (range, 91.7–112.7 s) for the CabTRP experiments. The frequency resolution of the spectrum is chosen by picking a value for the time-bandwidth product (denoted NW). Given a time-bandwidth product, the frequency resolution of the estimated spectrum is 2NW/T. The per-window spectrum estimate was calculated using NW = 4, yielding a frequency resolution of 0.05 Hz (range, 0.047–0.058 Hz) for the control experiments and 0.08 Hz (range, 0.07–0.09 Hz) for the CabTRP experiments. Seven tapers were used, and the estimates for each taper were averaged together within each window. The amount of window overlap was 50%, so that the last half of one window was the same as the first half of the next.
All spectrograms were normalized as follows. For each window, the total power was computed by integrating the power across frequencies. The average of the total power was then taken (across windows). All points of the spectrogram were divided by the mean total power to yield the normalized spectrogram. The raw spectrogram has units of Hz2/Hz whereas the normalized spectrogram has units of 1/Hz. The normed spectrogram was converted to decibels by taking the logarithm (base 10) of the normalized spectrogram and multiplying by 10. The normalized spectrogram in decibels (dB) was plotted.
Figure 1, B–F, shows spectrograms calculated from the activity of the LP, PD, DG, MG, and LG neurons shown in Fig. 1A. Power is represented by color, where colors toward the blue end of the spectrum indicate low power, and colors toward the red end of the spectrum indicate high power. The red areas in the spectrograms of the LP and PD neurons are at 0.94 Hz, the frequency of the pyloric rhythm (Fig. 1, B and C), whereas the red areas in the spectrograms of the DG, MG, and LG neurons are at 0.11 Hz, the frequency of the gastric mill rhythm (Fig. 1, D–F).
To quantify the amount of power in the pyloric and gastric frequency ranges (red and green brackets on the spectrograms in Fig. 1, C and F), we first plotted the amplitude (square root of power) of the total activity (black line) and the amplitude of activity in the gastric (green line) and pyloric (red line) frequency ranges (Fig. 1, B–F, 1st plot below each spectrogram). We normalized the amplitude in each frequency range to the total amplitude and plotted these over time (Fig. 1, B–F, 2nd plots below each spectrogram). In these plots, the x-coordinate of each point is the center of the window for which it was calculated, so the first and last points are one half the window duration away from the edges of the plot. CIs on these estimates were calculated using the jacknife estimate of SE, by performing take-away-one estimates of the logarithm (base 10) of amplitude and normed amplitude and using the percentiles of a t disribution with ntapers – 1 degrees of freedom to calculate an approximate 68% CI (Thomson and Chave 1991).
In the power spectra, rhythmic activity appeared as a peak at the fundamental frequency of the rhythm and also as harmonics at integer multiples of this frequency. Pyloric and gastric frequency ranges were chosen by eye to include the peak power at the fundamental frequency and to exclude the harmonics. To calculate the pyloric frequency, we first estimated an overall power spectrum for each spectrogram (data not shown) by averaging together the power spectra for every other window used in the spectrogram (so that the windows were not overlapping). To determine the frequency of the fundamental, we found the frequency with maximum power density from the relevant frequency range.
Coherence
To calculate the extent to which spiking activity in two signals was correlated as a function of frequency, we divided the data into the same number of temporal windows as for calculating spectrograms and estimated the normalized cross-spectrum (called the coherence) of the two signals in each window. The coherence is a complex-valued quantity, having both a magnitude and a phase at each frequency. The magnitude reflects the extent to which the variations in the two signals are correlated at the given frequency, and the phase reflects the extent to which one leads or lags the other. The coherence magnitude can take on values between 0 and 1, 0 indicating no correlation and 1 indicating perfect correlation.
Estimation of coherence was also done using the multitaper technique, with the same window lengths, time-bandwidth product, and number of tapers as above. The coherence magnitude was considered significantly different from 0 when it exceeded 
= 0.05 (Jarvis and Mitra 2001).
The coherograms in Fig. 1, G–I, show the frequencies at which there is coherent activity in the two indicated neurons and the phase relationship of that coherent activity over time. The coherograms show the coherence as a function of both time (x-axis) and frequency (y-axis), with coherence coded by color. Bright colors represent high coherence magnitude, and the particular hue indicates the phase of the coherence. Dark colors indicate low coherence. In the legend, the coherence phase is given as a fraction of a cycle, with a phase of 0 indicating synchronous activity and a phase of 0.5 indicating alternating activity. When calculating coherence, one signal is used as a reference, and the phase is that of the other signal (the "query" signal) relative to the reference signal. Thus a coherence phase of 0.25 means that the peak of the query signal occurred a quarter cycle after that of the reference signal (at the frequency in question). When describing a particular coherence estimate, we describe the estimate as query/reference. For instance, Fig. 1G shows a coherogram of DG and LP, with DG used as the query and LP used as the reference (thus it is labeled DG/LP, not LP/DG).
Estimation of coherence phase for whole traces
To estimate the relative phase of cells displaying pyloric activity, we used the phase of the coherence between the two cells. We collapsed the coherogram into a single coherence spectrum by averaging the complex coherence spectra for each window used in the coherogram. (To be precise, we averaged the coherence spectra from every other window, so that the windows were not overlapping.) Using the mean coherence spectrum (data not shown), we identified the frequency within the pyloric band that had the maximum coherence magnitude. We call this the peak coherence phase and use it as an estimate of the phase of one cell relative to the other for the pyloric-like activity. These estimates were in qualitative agreement with visual inspection of the data traces.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Figure 2 provides example recordings from adult LP, PD, and DG neurons (Fig. 2A) and from muscles innervated by those neurons—lpm, pdm, and dgm—in an embryo (Fig. 2B). In the adult, the LP and PD neurons are rhythmically active in the pyloric rhythm, and the DG neuron is active in the slower gastric mill rhythm. Note that, although the DG neuron is firing in long, slow, gastric-timed bursts, its activity is also modulated by the faster pyloric rhythm (Bucher et al. 2006). This is seen as the pyloric-timed interruptions during the DG neuron firing, as well as pyloric-timed subthreshold activity when the DG neuron is silent. In these data, as is typical in the adult, there is a clearly defined pyloric rhythm and a clearly defined and separate gastric mill rhythm.
|
), one that is slower and less regular than the adult pyloric frequency (Richards et al. 1999). Consistent with these results, the common rhythm in embryonic lpm and pdm was slower than the adult pyloric rhythm. In contrast to previous reports (Casasnovas and Meyrand 1995; Ducret et al. 2007; Le Feuvre et al. 1999), we reliably observed pauses in the dgm recordings, in which dgm failed to fire in time with the bursts in lpm and pdm (Fig. 2B). These pauses were reminiscent of the gastric-timed DG pauses observed in the adult (Fig. 2A). This suggests that these pauses might be precursor of the adult gastric mill rhythm, and the common embryonic rhythm might be a precursor of the adult pyloric rhythm. In contrast with the adult, however, the gastric-like activity did not always have one regular, clearly defined frequency; thus we call this "embryonic-slow activity." Similarly, to distinguish the common embryonic rhythm from the adult pyloric rhythm, we call it the "embryonic-fast rhythm" (Fig. 2B). Spectral analysis of adult motor patterns
The pyloric and gastric mill rhythms in adult lobsters and crabs have been extensively analyzed using traditional burst analyses that measure the period of the rhythm, the relative phase of firing of each neuron, and the duty cycle of each neuron (Bucher et al. 2005; Selverston and Moulins 1987). Embryonic motor patterns are often irregular, however, making burst detection difficult or impossible (Richards et al. 1999). Consequently, we used spectral techniques that do not depend on burst detection to analyze both adult and embryonic motor patterns in both control saline and in the presence of the neuropeptide CabTRP.
Figure 1 shows the use of spectrograms and coherograms to describe the activity of adult pyloric and gastric network neurons. Figure 1A shows the extracellular traces used to calculate the spectrograms. Spectrograms of activity in adult LP and PD neurons displayed high power at the pyloric frequency and low power at the gastric mill frequency, and spectrograms of activity in adult DG, MG, and LG neurons displayed high power at the gastric mill frequency and low power at the pyloric frequency (Fig. 1, B–F, top subpanels). These features are reflected in both the raw and normed amplitude in the pyloric and gastric mill bands: pyloric neurons have high pyloric amplitude and low gastric mill amplitude, and gastric mill neurons have high gastric mill amplitude and low pyloric amplitude (Fig. 1, B–F, bottom subpanels). Rhythmic activity in the adult was regular and stable over time; the high-power bands in all of the spectrograms (Fig. 1, B–F, top subpanel) are horizontal, and the power remains constant along the band. Similarly, the amplitude lines in the graphs below each spectrogram are all horizontal (Fig. 1, B–F, bottom subpanels).
Coherograms provide information about the relative phase at which different neurons are active. The brightest band in the DG/LP coherogram (Fig. 1G) is a narrow horizontal band at the pyloric frequency, indicating that the DG neuron exhibits a stable phase relationship with the LP neuron at the pyloric frequency. The LP and DG neurons burst in alternation and have a peak coherence phase of 0.5, consistent with the raw traces of LP and DG neuron activity (Fig. 1A). There is also stable coherent activity between the LP and MG neurons (Fig. 1H) and the LP and LG neurons (Fig. 1I), as shown by the horizontal bands at the pyloric frequency in these coherograms. These bands are at peak coherence phases of 0.1 for MG/LP and 0.2 for LG/LP, indicating that pyloric-timed activity in MG and LG occurs almost synchronously with LP bursting. These phase relationships are consistent with the apparent phasing seen in the traces and with traditional methods of estimating the phase relationships.
Pyloric neurons can also have gastric-timed activity, a phenomenon reported in the adult STNS (Bucher et al. 2006; Mulloney 1977; Weimann et al. 1991). This slow activity, seen dimly at 0.1 Hz in the LP spectrogram (Fig. 1B), is coordinated with the slow activity in MG (Fig. 1H) and LG (Fig. 1I). The low-frequency bands in the MG/LP and LG/LP coherograms are not as bright and straight and have more color changes than the higher frequency bands, indicating that, at least in this example, the slow coherent MG/LP and LG/LP activity is less stable than the fast coordinated activity (Fig. 1, H and I).
Fast and slow components of the embryonic motor pattern
Unlike adult rhythms that can be stably expressed for long periods of time (Fig. 1A), embryonic preparations often showed considerable activity variation, as seen in the recordings from embryonic lpm, pdm, and dgm (Fig. 3 A). Furthermore, although activity in embryonic lpm, pdm, and dgm appear to have the same burst frequency for some periods of time (Fig. 3A, 2nd inset), there were intermittent pauses in dgm activity that were not apparent in the recordings from lpm and pdm (Fig. 3A, 1st inset).
|
0.3 Hz). However, there is also significant power at lower frequencies in the lpm spectrogram (<0.14 Hz), arising from the pauses of different lengths that occurred in the dgm recording (Fig. 3D, within green bracket). For all three cells, the normed amplitude in the embryonic-fast rhythm frequency range is higher than that in the embryonic-slow activity frequency range (Fig. 3, B–D, bottom graph in each panel). However, the mean normed amplitude in the embryonic-slow frequency range is more than twofold higher in dgm (0.276 ± 0.033) than in lpm (0.137 ± 0.019) and pdm (0.114 ± 0.005). This embryonic-slow activity in dgm is less stable than activity in the higher frequency range; the embryonic-fast band in the dgm spectrogram has higher power, is narrower (even when viewed on a linear scale) and is more nearly horizontal (Fig. 3D). In all preparations (n = 10) in which simultaneous lpm, pdm, and dgm recordings were obtained, there was more slow activity in dgm as compared with the lpm and pdm. This is shown in the spectrograms of simultaneous lpm and dgm recordings from four representative preparations (Fig. 4). In all dgm/lpm experiments, there was slow activity in the dgm spectrogram that was not present in the lpm spectrogram from the same preparation. This slow activity appeared as a bright, low-frequency band that was clearly separate from the fast activity band in some of the dgm spectrograms (Fig. 4A, traces in Fig. 2B, and in three additional preparations; data not shown). However, in most cases, the pauses in dgm were quite irregular. This appears in the spectrograms as diffuse power at frequencies below the embryonic-fast frequency (Fig. 4, B and C). In still other preparations, slow dgm activity was minimal (Fig. 4D and two additional preparations; data not shown). In all dgm recordings, the fast activity was more stable over time than the slow activity (Fig. 4, A–D). Nevertheless, the majority of embryonic dgm recordings displayed some amount of embryonic-slow activity.
|
Activity expressed by other gastric mill neurons
Embryonic mgm also showed substantial embryonic-slow activity in five of five of preparations. The pauses in mgm activity were more pronounced than those in dgm (Fig. 5 A). As with dgm, these pauses occurred intermittently, with no clear temporal pattern. The spectrograms for lpm and mgm both show a band of activity at 0.5 Hz, corresponding to the embryonic-fast rhythm (Fig. 5, B and C). Additionally, the mgm spectrogram has substantial power at frequencies <0.3 Hz (Fig. 5, B and C). This low-frequency activity is separate from the embryonic-fast band and has peak power as high as that of the embryonic-fast band (Fig. 5C). The mean normed amplitude of the embryonic-slow activity in mgm (0.406 ± 0.067) is more than two times higher than that in lpm (0.181 ± 0.030). Overall, mgm activity is less stable than lpm activity—there are more variations in power in the mgm spectrogram compared with the lpm spectrogram.
|
|
Figure 7 shows one example recording from lgm. In this preparation, pauses in lgm activity were rare (Fig. 7A). Consistent with this, the spectrograms show that the frequency content of lpm and lgm was similar over the length of these recordings (Fig. 7, B and C). The power at all frequencies fluctuated more in lgm than in lpm (Fig. 7, B and C), suggesting that lgm activity was less stable than lpm activity in this preparation.
|
0.5 Hz) predominated in the spectrograms of lpm and lgm recordings from four different preparations (Fig. 8, A–D). The embryonic-fast rhythm was strong and stable over time, whereas embryonic-slow activity was weak. In three of these four preparations, activity in dgm was also recorded (data not shown). In all of these preparations, there appeared to be more slow activity in the dgm spectrograms compared with the lgm spectrograms, although these differences were not significant (unpaired t-test, P = 0.32; mean normed slow amplitude lgm = 0.23 ± 0.06; mean normed slow amplitude dgm = 0.31 ± 0.13).
|
0.5 Hz). This coordinated fast activity was stable (Fig. 8, A–D). The embryonic lgm and lpm fired synchronously at the common embryonic-fast frequency (Fig. 8, A, C, and D). There was little coordinated slow activity in the lgm/lpm coherograms, and whatever coherence was present had no clearly defined phase relationship (Fig. 8, A–D). Summary of embryonic and adult motor patterns
Figure 9 shows the summary of the mean normed amplitudes of the embryonic-fast and embryonic-slow activity in all recordings. The majority of the power in the output from all neurons was in the embryonic-fast frequency range regardless of cell type. Nevertheless, there was significantly more embryonic-slow activity in both dgm and mgm as compared with lpm and pdm (Fig. 9, statistics in figure legend). lgm appeared to have more embryonic-slow activity than lpm and pdm and less than both mgm and dgm, but these differences were not statistically significant (Fig. 9, statistics in figure legend).
|
We wanted to quantify the difference in frequency between the adult pyloric rhythm and the embryonic-fast activity. Using the peak frequency of lpm as a measure of the embryonic-fast frequency (see METHODS), we found that the mean embryonic-fast frequency (0.50 ± 0.21 Hz, n = 25) was significantly slower than the mean adult pyloric frequency (0.81 ± 0.10, n = 14; unpaired t-test, P < 0.0001). This is consistent with previous reports, which established that the embryonic-fast activity was slower than the adult pyloric rhythm (Richards et al. 1999).
We were curious to know how the phase relationships of the embryonic-fast activity compared with those of the adult pyloric rhythm. The phase relationships of the embryonic-fast activity and the adult pyloric rhythm, as assessed by the peak coherence, are shown in Fig. 10. In both embryos and adults, PD activity alternated with LP (Fig. 10, A and B). In all of the embryonic gastric mill neurons, the activity occurred in the later part of the lpm cycle, but activity in both lgm and mgm was closer to firing synchronously with lpm, and dgm activity was closer to alternating with lpm (Fig. 10A). Pyloric-timed activity in adult DG and LG were significantly earlier in the LP cycle than in the embryonic-fast rhythm (statistics in figure legend). Thus PD and MG phasing was similar in the adult pyloric rhythm and the embryonic-fast activity, but LG and DG phasing was different.
|
0, adults at 0.25). CabTRP effects on adult motor patterns
We have shown that embryonic dgm and mgm display significant embryonic-slow activity. Although this activity is likely a precursor to the adult gastric mill rhythm (see DISCUSSION), it is weaker and more irregular than the adult gastric mill rhythm. Why might this be? One possible explanation is that, in the adult, certain neuromodulators are supplied by descending inputs to the STG, and these modulators are diminished or absent in the embryo. Previous studies in the adult STG have shown that neuromodulators can alter the extent to which neurons participate in the pyloric or gastric mill motor patterns (Marder and Bucher 2007). One modulator, CabTRP, can increase slow activity in gastric mill neurons of the crab, C. borealis. Furthermore, CabTRP is absent from the descending inputs in the embryonic STG but is present in the adult (Cape et al. 2008; Fénelon et al. 1999). Therefore the relative weakness of the slow activity in embryonic gastric mill neurons could be explained by the absence of CabTRP from the embryonic inputs (Cape et al. 2008; Fénelon et al. 1999).
We first examined the effects of CabTRP on the gastroplyoric motor patterns of the adult lobster. CabTRP did not have obvious effects on LP and PD neurons, but it eliminated pyloric-timed activity in DG, allowing the gastric mill rhythm to predominate (Fig. 11, A and B). Figure 12 shows the spectrograms of the traces in Fig. 11. In control conditions and in CabTRP, most of the activity in LP and PD occurred around 0.8 Hz (Fig. 12, A and B). In control, the DG neuron showed two very different frequencies in its activity: a pyloric-timed activity (0.8 Hz) and a gastric-timed activity (<0.2 Hz; Fig. 12A, right). In CabTRP, the pyloric activity disappeared, and most of the DG activity occurred at the gastric mill frequency (Fig. 12B, right).
|
|
In the embryo, CabTRP increased the amount of embryonic-slow activity in dgm, thus transforming the embryonic motor pattern into a more adult-like form. Figure 13 shows simultaneous intracellular recordings of activity in lpm, pdm, and dgm in one embryo preparation. In control saline, the frequency of lpm, pdm, and dgm activity was 0.3 Hz (Fig. 13A). When CabTRP was bath applied, the frequency of lpm and pdm activity increased to 0.42 Hz, but the most prominent effect of CabTRP was on dgm, in which CabTRP produced intermittent pauses (Fig. 13B).
|
|
The effects of CabTRP on the amount of slow and fast activity were similar, although not identical, in embryos and adults (Fig. 15). CabTRP significantly increased the amount of slow activity in both adult DG and embryonic dgm (Fig. 15, A–D; statistics in figure legend). CabTRP also significantly increased the amount of slow activity in adult LP and PD neurons and decreased the amount of fast activity in adult LP and DG neurons (Fig. 15, A and B; statistics in figure legend).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Spectrograms to measure fast and slow activity
To quantify the amount of fast and slow activity in the embryonic recordings, we calculated spectrograms, a method commonly used for acoustical, thermal, seismic, and electromagnetic signals (Thomson 2000). Biologists have used spectrograms to analyze a wide variety of phenomena, including acoustic communication (e.g., birdsong) (Margoliash 1997; Marler 1969), behavioral and circadian oscillations (Dolezelova et al. 2007; Dowse 2007; Levine et al. 2002; Sanyal et al. 2006), and rhythmic activity in neuronal networks (Bucher et al. 2006; Miller and Sigvardt 1998; Richards et al. 1999). These tools, although not commonly used for analyses of central pattern generating networks, will be invaluable for the study of motor patterns and other rhythms, especially in genetically tractable model organisms. In such animals, genetic manipulations commonly generate erratic and irregular neuronal and behavioral phenotypes, which nonetheless must still be assayed for rhythmicity (Dixit et al. 2008; Dolezelova et al. 2007; Ebert et al. 2005; Fox et al. 2006; Kopp et al. 2005; Nitabach et al. 2006; Sanyal et al. 2006).
We used spectrograms because traditional methods of analysis require unambiguous burst detection, which was not possible for many embryonic recordings, and because spectral analysis was used in a previous study to quantify the relative amount of slow and fast activity in the adult lobster STNS (Bucher et al. 2006). In this study, we analyzed relatively long stretches of data (10–110 min). These long recordings enabled us to show for the first time that embryonic gastric mill neurons produced significantly more embryonic-slow activity than pyloric neurons and that the frequency of both the fast and slow activity varied across the recordings. This lack of stationarity could be caused by changes in activity in descending modulatory inputs, a result of sensory feedback, or result from networks that are less tightly tuned in the adult, and that may be intrinsically "noisy" in their output. This kind of intermittent output is reminiscent of the "bouting" seen in preparations of the crab STG that recover rhythmicity after the removal of modulatory inputs (Khorkova and Golowasch 2007; Luther et al. 2003), presumably as the network retunes itself.
Coherograms to estimate phase
The phase relationships of activity in stomatogastric neurons are typically defined as the relative onset and offset of each burst (Selverston and Moulins 1987). Although both we and Le Feuvre et al. (1999) recorded short sequences of data in which conventional burst analyses was feasible, the lack of stationarity in long recordings in which irregular sequences were interspersed with more regular stretches precluded its use in all but selected short stretches of data. In contrast to burst detection methods, coherograms allowed us to calculate the stability of the phase relationships over extended time periods. The coherogram-derived phase relationships of embryonic-fast activity were stable and similar to those determined by burst onset/offset (Le Feuvre et al. 1999), but the phase relationships of the slow activity were quite variable. When comparing adults and embryos, we found that phase relationships of the output from embryonic and adult PD/LP and MG/LP neurons were not statistically different, and LG/LP and DG/LP phases were different across development.
Comparison of embryonic and adult activity
We suggest that the embryonic-fast activity in pyloric neurons is an immature form of the adult pyloric rhythm because 1) the output from embryonic LP and PD neurons fire out of phase, as they do in the adult, and 2) during larval development, embryonic-fast activity from LP neurons becomes faster and more regular, becoming more adult-like by the end of larval stage (Richards et al. 1999). Fast activity was also produced by embryonic gastric mill neurons, as it is by adult gastric mill neurons, particularly DG neurons when the gastric mill rhythm is present (Bucher et al. 2006). However, the relative phase of this fast activity in LG and DG neurons was different in embryos and adults. Modeling work (Bem et al. 2002) suggests that electrical coupling, which is believed to be higher in embryonic STG (Ducret et al. 2006, 2007), could be responsible for a greater degree of synchrony, perhaps contributing to the developmental differences of the LG and DG phase relationships.
It is tempting to propose that the embryonic-slow activity is an immature form of the adult gastric mill rhythm. Evidence in favor of this hypothesis is the observation that embryonic-slow activity was reliably present only in gastric mill neurons. Furthermore, in a small set of experiments, the frequency of the embryonic-slow activity was similar in all monitored gastric mill neurons from the same embryonic preparation. On the other hand, the phase of this slow activity was not identical in embryos and adults. However, the phase of the gastric mill neurons can vary considerably depending on how the gastric mill rhythms are elicited (Blitz et al. 1999; Combes et al. 1999a,1999b; Heinzel and Selverston 1988; Norris et al. 1994). Therefore it is difficult to know whether the differences in phase are indications of different forms of "gastric-like" activity or caused by a fundamental immaturity of the network.
Effects of CabTRP on adult STNS motor patterns
In the adult STG, neurons can simultaneously exhibit both fast and slow rhythmic activity, to varying degrees (Bucher et al. 2006; Weimann and Marder 1994; Weimann et al. 1991). Sensory and neuromodulatory input can influence the amount of fast and slow activity by altering neuronal excitability and/or synaptic strengths (Hooper and Moulins 1989; Weimann et al. 1990, 1993). In the gastric mill neurons of the crab C. borealis, CabTRP enhances slow activity, presumably by increasing the excitability of the gastric neurons, changing synaptic strengths between the gastric neurons, or decreasing the synaptic strengths from pyloric neurons onto gastric mill neurons (Wood et al. 2000). In the adult lobster, CabTRP strongly enhanced gastric bursting. In fact, the decreased pyloric-timed activity seen in the DG neuron (Fig. 11) might indicate that CabTRP is
